def sample_user_pos(user_inds_batch: Collection[int],
pos_xn_csr: sp.csr_matrix, rand: np.array, *_):
batch_size = len(user_inds_batch)
rand_rats = rand.rand(batch_size)
pos_l = []
for user_ind, rat in zip(user_inds_batch, rand_rats):
# TODO: `get_row_nz` repeated in sampling
user_pos_item_inds = get_row_nz(pos_xn_csr, user_ind)
user_pos_item = user_pos_item_inds[int(rat * len(user_pos_item_inds))]
pos_l.append(user_pos_item)
pos_item_inds_batch = np.array(pos_l)
return pos_item_inds_batch
def sample_user_pos_weighted(
user_inds_batch: Collection[int],
pos_xn_csr: sp.csr_matrix,
rand: np.array,
cs_l: List[np.array],
):
batch_size = len(user_inds_batch)
rand_rats = rand.rand(batch_size)
pos_l = []
for user_ind, rat in zip(user_inds_batch, rand_rats):
cs = cs_l[user_ind]
pos_item_inds = get_row_nz(pos_xn_csr, user_ind)
ind = np.searchsorted(cs, rat, side='right')
user_pos_item = pos_item_inds[ind]
pos_l.append(user_pos_item)
pos_item_inds_batch = np.array(pos_l)
return pos_item_inds_batch
